Patient Spontaneous Effort Estimation in Digital Twin Model with B-spline Function

Patient work of breathing is of significant clinical interest for several decades. It is particularly relevant when gauging patient-ventilator interaction, patient-specific level of mechanical ventilation (MV) support, and weaning from full support to spontaneous breathing MV modes or off of MV entirely. Current monitoring approaches require additional equipment (usually expensive), well-trained clinicians (to collect and interpret these signals), or/and extra clinical interventions (increases difficulty and cost). This study extends a digital twin model to estimate patient spontaneous breathing effort (PPPP curve) with a previously proposed model-based estimation method using b-spline functions. Data from 22 patients for two assisted MV modes, NAVA (neurally adjusted ventilatory assist) and PSV (pressure support ventilation), are employed. Estimation results are compared to breathing effort reflected by electrical activity of the diaphragm (EAdi) signals. Physiologically- relevant correlations in identified PPPP curve area (negative and positive) and EAdi signal can be found in both NAVA and PSV data analysis. While PPPP curves yield more negative area (larger PRCTneg), the corresponding breaths tend to have lower peak EAdi values and area under curve of EAdi signal (AUC[EAdi]) during inspiration. R2 values for NAVA data yield an interquartile range (IQR) from 0.31 to 0.68 for peak EAdi versus PRCTneg and 0.40 to 0.61 for AUC[EAdi] versus PRCTneg, respectively. Results differ between NAVA and PSV modes based on poorer patient-ventilator interaction observed in PSV, while the same level of expected physiological relevance is still observed. Overall, the extended digital twin model with b-spline functions to quantify patient-specific inspiratory effort shows promising application in helping guide weaning or changes in MV settings at bedside for assisted breathing modes of MV. In future, the identified PPPP curves could also potentially be used to replace the need for costly measurement of EAdi signals. Copyright (c) 2023 The Authors.